E2E Networks Ltd Q1 FY2025 Concall Summary

Key Takeaways

• Revenue doubled, GPU workloads now dominant: Revenue grew 112% YoY and ~40% QoQ, with cloud GPU AI/ML workloads now constituting the majority of business.
• Margins surged on compute density: EBITDA margin reached 66.34%, up 1403 bps YoY, as the shift from low-density CPU to high-density GPU compute structurally lowers opex as a proportion of revenue.
• A specialized cloud niche is opening: Hyper scalers charge 3x-4x for GPU compute, creating space for providers like E2E that deliver similar functionality at half the price or lower.
• India's compute gap is the structural tailwind: India holds 20% of the world's data but only 2% of its compute capacity, implying a ~10x buildout opportunity, particularly in GPU infrastructure.
• What to watch next quarter: Whether the next tranche of 256 H100 GPUs deploys on schedule, how customer concentration normalizes after a bursty quarter, and any concrete progress on the three government business pathways now that MeitY empanelment is in hand.

The GPU Density Shift

• Economics of density: GPU servers cost 10x, occupy ~2x space, and consume ~10x power versus CPU servers, but deliver ~1000x compute for massively parallel workloads. This density means data center costs as a percentage of revenue keep falling as the GPU mix rises.
• Opex trajectory: Operating expenses dipped by a greater margin this quarter, and management expects the percentage of data center cost to revenue to continue declining as compute density increases.
• Platform leverage: The platform business model means team size does not need to scale with cloud revenue; the business can potentially scale 10x with the platform remaining practically the same size, leaving scope for margin growth.
• Margin sustainability: Management stated the medium-term and long-term broad trend should sustain, though it declined to confirm whether margins have moved from the +50% range to +60%.

Dua: "the worst of the scarcity is now behind us. So we should be able to kind of plan our hardware run for much shorter period now, compared to what we did in the past, when there was a lot of scarcity."

The Specialized Cloud Opportunity

• The pricing wedge: As GPU compute becomes a larger share of organizational expense, a space has opened for specialized cloud providers that are not as expensive as hyper scalers but deliver ~90% of what most organizations need while doing ~10% of the features.
• No 1:1 comparison: Management cautioned that there is never a 1:1 price comparison because of tremendous differences in features between providers, and declined to address competitor pricing directly when asked.
• International traction emerging: The company is already seeing some level of traction outside India, and management believes international customers could reach 25% to 30% of the AI business over a year, though it called any current-quarter figure premature to comment on.
• Data localization green shoots: Indian companies are beginning to recognize the need to work with Indian vendors, though management said it is not yet quantifiable whether this becomes a large trend.

Dua: "there is suddenly a space for specialized cloud providers who are not as expensive as hyper scalars. But in terms of capabilities, they provide almost a similar functionality, maybe doing 10% of the features, but delivering 90% of what is needed by most organizations."

Customer Dynamics

• ARPU surge: ARPU rose 48% QoQ, driven by larger customers growing their cloud GPU AI/ML workloads.
• Concentration spike: Top 10 customer concentration was higher than usual this quarter at 45% to 50%, because cloud GPU workloads are bursty: a single customer may have a short burst of large compute usage and be significantly lower the next quarter.
• Enterprise margin tradeoff: Solving deeper enterprise problems initially impacts margins during commissioning, but eventually results in much higher ARPU and higher margins due to stickiness and lifetime value.
• Customer vs investor focus: Management explicitly prioritizes customer conviction over investor conviction, stating it only needs to convince customers to use the product.

Dua: "I only want to convince our customers to use our product, I don't need to convince anyone else."

India's Compute Buildout

• The 10x gap: India has 20% of the world's data but only 2% of the world's compute capacity, measured by power consumed by compute systems in Indian data centers. Management sees this as a ~10x opportunity for compute infrastructure design and delivery.
• Bright future conviction: Management is very hopeful of a bright future for the cloud GPU industry in India, based on what is observed in developed markets today.
• Cooling timeline: High-density compute of the kind being planned is still about 1.5 to 2 years away from requiring different cooling facilities, so near-term infrastructure is manageable with current setups.

Dua: "India only has 2% of the world compute capacity in terms of the amount of power, which has consumed by our compute and the amount of compute systems we have running in our data centers. So which means that, about 10x, the amount of compute still needs to be kind of designed and delivered for Indian organization"

Government & Certification

• MeitY empanelment achieved: E2E has been certified by MeitY less than a month ago, signifying suitability of the cloud platform for government compute workloads.
• Three government pathways, all early: (1) Support MSME partners on the GEM portal on a case-by-case basis, very early given the recent certification; (2) evaluate where E2E's capabilities match government workloads over the next 18 to 24 months; (3) combine capabilities with partners. All three are far enough in the future to comment on today.
• No near-term quantification: Management characterized the government opportunity as something that will play out over the next 18 to 24 months, with nothing concrete to report now.

CAPEX & Balance Sheet

• No FY25 CAPEX commitment: Management explicitly walked back any characterization of an ₹800 crore CAPEX plan, stating that figure was an answer to "how much would you like to spend" and not an announcement. It could be more or less, and the company is very flexible.
• Funding mix: CAPEX will be funded through a judicious mix of vendor financing, debt, internal accruals, and potentially some equity raise, with any equity raise to be informed to the exchange before pursuing.
• Planning cycle shortened: Hardware planning cycles can now be drastically shortened to about 6 to 8 weeks, versus 6 to 12 months previously, now that the worst of hardware scarcity is behind.
• Asset turns: Broadly between 0.5 to 0.6, depending on which SKUs customers choose to deploy.
• Current deployment: The entire deployment of last year's ₹185 crore CAPEX was completed by end of March or first week of April, and overall inventory utilization is about +90%.
• Next GPU tranche: Already in the process of acquiring another 256 H100 GPUs, which would go live in the current quarter.

Dua: "I will not take that bait this time, and I will say that let's wait and watch. We will keep everyone informed of our plans as they get formalized."

Guidance & Commitments

Key Metrics